Development of a surrogate model of an amine scrubbing digital twin using machine learning methods

This research plan tackles the slow calculations that come with detailed chemical process simulations. It proposes building a machine‑learning surrogate model for an industrial amine‑scrubbing digital twin. We start with a steady‑state simulation in Aspen HYSYS that has been checked against real plant data. The goal is to create a fast, computationally cheap alternative that can be used for real‑time optimization and control. To generate the surrogate data, we use a structured design of experiments. Specifically, we apply Latin‑Hypercube sampling to create nested operating regions around the normal steady‑state point. Several machine‑learning regression methods are then trained and tested with strict cross‑validation. This lets us pick the best predictor for

each process variable. By swapping out heavy thermodynamic calculations for quick statistical models, the method enables fast scenario analysis, predictive maintenance, and real‑time decisions in carbon‑capture and gas‑purification units. The framework also gives detailed instructions on how to generate data, choose models, and validate them, providing a solid base for using digital twins in complex chemical processes.

Message: Hi FE, I need help with “Development of a surrogate model of an amine scrubbing digital twin using machine learning methods” in “Chemical Engineering”